Metraipolis › archive › dc_1682bec8ed55
Multi-step SWE: Can Coding Agents Keep Building on Their Own Code Over Long Horizons? — arXiv:2611.07166
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.07166
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-06T14:42:12.253Z
- SHA-256 of the archived text
1305f833a9bc263af3aeb222add5b4c5dcd8bf996362aa3685e9d9ddb4571989- Stable URL of this page
https://metraipolis.com/documento/dc_1682bec8ed55
How to cite
Multi-step SWE: Can Coding Agents Keep Building on Their Own Code Over Long Horizons? — arXiv:2611.07166 — arXiv (en), https://arxiv.org/abs/2611.07166 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_1682bec8ed55 (Metraipolis archive dc_1682bec8ed55, sha256 1305f833a9bc263af3aeb222add5b4c5dcd8bf996362aa3685e9d9ddb4571989)
BibTeX
@misc{dc1682bec8ed55,
title = {Multi-step SWE: Can Coding Agents Keep Building on Their Own Code Over Long Horizons? — arXiv:2611.07166},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.07166},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_1682bec8ed55; sha256 1305f833a9bc263af3aeb222add5b4c5dcd8bf996362aa3685e9d9ddb4571989},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.07166. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Developers rarely give a coding agent a single request: they ask it to build a component and then to extend, upgrade, and optimize it, each time on code the agent itself wrote. Most benchmarks, however, score one task at a time. We introduce Multi-step SWE, a long-horizon benchmark of 69 software engineering tasks, each a chain of 5-15 dependent steps on a single codebase. Every step is a complete, multi-hour engineering request that starts from the code the agent produced in the preceding steps, so early design decisions and errors carry forward. The tasks span four families of engineering work and nine languages, include both from-scratch builds and existing open-source codebases, and ground every step in an existing source of truth that defines both the instruction and its verifier. Across nine frontier models, the strongest model reaches a solve rate of only 18.5%, and 38 of the 69 tasks are not solved by any model. We release Multi-step SWE publicly at this https URL . Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Software Engineering (cs.SE) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2611.07166 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-11-06T14:18:42.667Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).